Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Quantum from principles
1) Pure states
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Quantum theory
3
Introduction
● Quantum theory can be defined via a short number of clear postulates.
● Although these postulates are strictly speaking independent, one can ask questions of the following type:
● “Are there theories with the same pure states as quantum theory?”
4
Introduction
● Need a general framework to talk about theories.● What type of theory is quantum theory?● Is it an information theory? A logic? A form of
mechanics? A general probability theory? A process theory?
5
Introduction
● Are there logics with the same measurement structure as quantum theory but different states?
● Are there operational theories with the same pure states and dynamics as quantum theory but different measurements?
● The answers to these questions will tell us how independent the postulate are.
6
Introduction
● Is it possible to change one part of quantum theory whilst keeping the others constant?
● Could we obtain a future theory of physics by making “perturbations” to the current theory? Or will we need a radical overhaul?
7
Introduction
● If we assume quantum theory is a form of logic, then the structure of measurements fully encodes the probability rule.
● Assuming it is a GPT means that the dynamical part fully determines the probabilistic part.
● Adopting different perspectives on quantum theory will allow for different insights into what it has to tell us.
● I don’t argue that any of the perspectives I adopt today are the final word, rather they are useful stepping stones.
Quantum logic and Gleason’s theorem
9
Mathematical Foundations of Quantum Mechanics, Von Neumann (1932)
10
11
Quantum logic
● Facts about quantum systems do not obey the rules of classical logic
● Need a new set of rules (a quantum logic) to reason about propositions
Classical logic
13
Classical logic (propositional)
● A set of rules about propositions● Propositions can be true or false● Can connect propositions together to form
formulae using connectives: “and” , “or”● One connective is implication
14
Classical logic (propositional)
● A set of rules about propositions● Propositions can be true or false
P Q P or Q
T T T
T F T
F T T
F F F
15
Classical logic
● A set of rules about propositions● Propositions can be true or false
16
Classical logic as set theory
p
17
Classical logic as set theory
p
p = being red = all red objects
18
Logical connectives in set theoretic language
p q
p = being red = all red objects
q = being square = all square objects
19
Logical connectives in set theoretic language
● “and” is set intersection
p q
p = being red = all red objects
q = being square = all square objects
20
Logical connectives in set theoretic language
● “and” is set intersection
p q
p and qp = being red = all red objects
q = being square = all square objects
21
Logical connectives in set theoretic language
● “or” is set union
p q
p or qp = being red = all red objects
q = being square = all square objects
22
Logical connectives in set theoretic language
● “or” is set union
p q
p or q
23
Classical logic as set theory
p = “particle is in the blue circle”
24
Logical connectives in set theoretic language
● “not” is complement set
not p = “particle is not in the blue circle”
25
Logical connectives in set theoretic language
● What is implication?● All things p are also q● “Being a square implies having four sides”● All squares are four sided
26
Implication
q
27
Implication
qp
28
Implication as set inclusion
● Implication is given by set inclusion
qp
29
Classical logic
● Classical physics:
a cb d
p = “particle is in interval [a,c]”
30
Classical logic
● Classical physics:
a b c d
q = “particle is in interval [b,d]”
31
Classical logic
● Classical physics:
a b c d
p and q = “particle is in interval [b,c]”
32
A simple example
33
A simple example
34
A simple example
35
Logic as set theory
● Propositions correspond to sets.● Implication amongst propositions given by set
inclusion.● Set inclusion is a partial order:
36
Implication
qp
37
Implication
qp
38
Lattice
39
Complete lattice
● Complete lattice: Partially ordered set where every pair of elements has a “and” and “or” defined (meet and join)
● Meet: greatest lower bound● Join: least upper bound
40
Complete lattices
41
Complete lattices
Meet? Greatest lower bound
42
Complete lattices
Meet? Greatest lower bound
43
Complete lattices
Meet? Greatest lower bound
44
Complete lattices
Join? Lowest upper bound
45
● A logic is just a complete lattice● Classical logic is distributive lattice● p and (q or r) = (p and q) or (p and r)● p or (q and r) = (p or q) and (p or r)
46
● A logic is just a complete lattice● Classical logic is distributive lattice● Quantum logic is a non-distributive lattice● We have a vantage point to talk about general
non-classical logics, of which quantum logic is an example.
47
Quantum logic
● Propositions correspond to subspaces of H.● Examples: , ,● Quantum logic is a lattice of subspaces
48
Quantum logic
49
Quantum logic
50
Quantum logic
51
Quantum logic
● Propositions correspond to subspaces of H.● Implication given by inclusion of subspaces
52
Quantum logic
53
Quantum logic
● Propositions correspond to subspaces of H.● “and” is given by intersection
54
Quantum logic
55
Quantum logic
● Propositions correspond to subspaces of H.● “or” is given by span
56
Quantum logic
57
Quantum logic
58
Quantum logic
● Propositions correspond to subspaces of H.● Negation is given by orthogonal subspace
59
Quantum logic
60
Quantum logic
● Propositions correspond to subspaces of H.● Implication given by inclusion of subspaces● “and” is given by intersection ● “or” is given by span● Negation is given by orthogonal complement
(orthocomplemented lattice)
61
Quantum logic
62
Non-distributivity of quantum logic
63
Non-distributivity of quantum logic
64
Non-distributivity of quantum logic
65
Non-distributivity of quantum logic
66
Non-distributivity of quantum logic
67
Non-distributivity of quantum logic
68
Non-distributivity of quantum logic
69
Non-distributivity of quantum logic
70
Operational quantum logic
● Propositions (subspaces) can be thought of as answers to yes/no questions (Mackey).
● These questions are the (projective) measurements we carry out.
71
Gleason’s theorem
● If the answers correspond to subspaces, what are the possible probability assignments?
● Measure
● For mutually exclusive answers we expect probabilities to add:
72
Gleason’s theorem
● If the answers correspond to subspaces, what are the possible probability assignments?
● For projectors onto orthogonal subspaces (mutually exclusive properties)
73
Gleason’s theorem
The only measures on L(H) (for ) which obey the lattice structure are of the form
Where is a density operator
74
Axiomatisations of quantum theory
● Assume the structure of questions and answers● The only probability rule compatible with the
lattice structure of subspaces is the Born rule● Recover the probability rule and states.
75
A sketch for a reconstruction
● Assume structure of questions● Use Gleason’s theorem to get states and
probability rule● Use theorem by Wigner to get the unitary group
76
Conclusion
● We adopted a certain perspective on quantum theory.
● Namely it is an Operational Quantum Logic● Using this perspective we see that the axioms
about the probability rule (and states) directly follow from the structure of projective measurements.
77
Conclusion
● In classical physics experimental propositions can be identified with sets
● In quantum physics experimental propositions can be identified with subspaces
● Both have a lattice structure
78
Where next?
● We will adopt a different operational perspective● That of General Probabilistic Theories● Using this perspective we see that the axioms
about the probability rule (and measurements) directly follow from the structure of pure states.
General probabilistic theories
80
Operationalism (1927)
https://www.pitt.edu/~jdnorton/teaching/HPS_0410/chapters/significance_2/index.html
81
Operationalism
● “We mean by any concept nothing more than a set of operations; the concept is synonymous with the corresponding set of operations.” Bridgman (1927)
● For instance concept of length should reduce to a set of operations one can carry out with a ruler.
82
Operationalism
● Special relativity● Theories defined via a small number of
operational principles.● Operational reconstructions of quantum theory● See Lluis Masanes lectures from last summer
school
83
General probabilistic theories
84
The operational approach to physical theories
● In the operational approach a theory just allows us to make predictions about the outcomes of measurements.
● No claims are made about ontology or underlying physical reality.
● We have access to some physical devices which we can wire together, and the theory allows us to make predictions about what will happen.
85
The operational approach to physical theories
● Operational principles will lead to certain mathematical structure
● Mixing: convex structure● Composition of devices: categorical structure
86
Primitives
● The primitives are devices: preparation, transformation and measurement devices.
87
The operational approach
● The primitives are devices: preparation, transformation and measurement devices.
Preparation choice
88
Measurement device
● The primitives are devices: preparation, transformation and measurement devices.
Measurement readoutMeasurement choice
89
Prepare and measure
● The primitives are devices: preparation, transformation and measurement devices.
Preparation choice Measurement readoutMeasurement choice
90
An example: qubit
Prepare Outcome Measure in Z basis
91
Transformation devices
92
The operational approach in quantum foundations
● For various different settings of the preparation device we collect the statistics for various measurements.
● We use these statistics to generate the state space and effect space.
93
An example: qubit
Prepare Outcome Measure in Z basis
Outcome
94
The operational approach in quantum foundations
● The classical world: laboratories, agents and classical probability theory is assumed.
● Fundamentally not about systems, but about devices, and the correlations between classical inputs and outputs of these devices.
● From the collected statistics we derive the state and effect space.
95
Classical probability theory (finite dimensional)
wikipedia Getty images
96
Classical systems (finite dimensions)
● Consider a ball which can be placed in one of two boxes:
97
Classical systems (finite dimensions)
98
Classical systems (finite dimensions)
99
Preparations
● A preparation consists of putting the ball in one of the boxes and closing them.
● I can choose to put the ball in box 1 with some probability p, and box 2 with probability 1-p
p
1-p
100
Preparations
● A preparation consists of putting the ball in one of the boxes and closing them.
● I can choose to put the ball in box 1 with some probability p, and box 2 with probability 1-p
● This is called preparing an ensemble, and is a central assumption of the GPT framework
101
Preparing an ensemble
p
1-p
102
Measurements
● I give the closed box to someone else● They can carry out a measurement: open the box
to see which partition the ball is in● For a given preparation we repeat this many
times● The measurement has two outcomes: “The ball is
in box 0” and “The ball is in box 1”.
103
An example
Prepare
Outcome Measurement: open both boxes
104
An example
Prepare Outcome
Measurement: open both boxes p
1-p
Outcome
105
States
● For a given preparation outcome 0 will occur with probability p, and outcome 1 with probability 1-p
State just allows us to make predictions.
106
States
p
1-p
107
States
● A state is a probability distribution over outcomes
● Just like in operational quantum logic
108
State space
109
State space
Extremal states (pure states)
110
State space
Convex hull of extremal points
111
Convex combination
112
State space
Extremal states (pure states)
Mixed states (describes an ensemble)
113
Subnormalised states
● The experimenter can also choose to just send an empty box
●
●
● This will give probability 0 for all measurement outcomes, and hence is represented by
●
114
115
Cone of states
Normalised states
Subnormalised states
116
Effects
● How do I obtain the probability p(0|P) of obtaining outcome 0 from the state ?
117
Effects
What is probability of outcome 0 or outcome 1 occurring?
What is probability of neither outcome occurring?
118
Subnormalised states
● The experimenter can also prepare mixtures of the empty box and preparations with a ball in one of the partition
p
1-p
119
Effects
is an effect. It is a linear functional of states (dual vector).
120
121
Effect space
122
Effect space
123
Effect space
124
The trit
125
The trit
126
Trit state space
127
Classical four level systems
● Tetrahedron, embedded in , just drawing the normalised states
128
Classical systems
● A classical system with d pure states is a simplex with d extremal points.
● Line segment, triangle, tetrahedron...
129
Deriving the quantum state space
130
Qubit: deriving the state space
● Pure states● Two outcome measurements
● Can prepare ensembles
131
Qubit: deriving the state space
0.25
0.75
132
Qubit
Could write a state as a vector of probability distributions over outcomes
●
133
Qubit
● Could write a state as a vector of probability distributions over outcomes.
● But would be infinitely long.● And contains redundancies.
●
134
Qubit
● Any can be written as a linear combination of 4 functions of the form
● For instance we can have the +X,+Y,+Z and -Z outcome probabilities as the basis functions.
● Could be a SIC-POVM
135
The probabilistic representation
136
The probabilistic representation
Any outcome probability can be computed as a linear function of the 4 fiducial probabilities.
137
Representing ensembles
is the mixed state representing the ensemble P
138
Representing ensembles
● Mixed state representing ensemble is obtained by taking convex combination of vectors
139
Qubit state space
https://commons.wikimedia.org/w/index.php?curid=5829358
140
Probabilistic representation
● This is the representation of states in which ensembles are represented by convex combinations of state vectors.
● For example the Bloch vector representation/density matrix representation.
● This is the representation we use in GPTs.
141
Key facts about the probabilistic representation
● Outcome probabilities are linear functions of the states.
● We went from to ● We can take mixtures of states by taking convex
combinations of state vectors.● Since for any pair of states we can take a mixture
the state space is convex.
142
Convex sets
● For every pair of points in the set the line segment joining them is part of the set.
143
● Classical and quantum state spaces correspond to convex sets.
● Other convex sets correspond to other non-classical systems.
● Like the general lattice structure and non-classical logics.
144
General GPT: single system
State space
Effect space
145
How to “read” a convex state space
Mixtures of states:●
a b
c
d
0.25
0.75
0.5
0.5
a
b
c
d
146
“True” definition of mixed states
● Multiple ensembles are assigned the same mixed state.
● A mixed state is an equivalence class of ensembles.
● See Holevo: “Probabilistic and Statistical Aspects of Quantum Theory”
147
Perfectly distinguishable states
a
b
Outcome 0
Outcome 1
148
How to “read” a convex state space
● Effects are linear functionals.
149
How to “read” a convex state space
● Effects are linear functionals.
150
Perfectly distinguishable states
● So antipodal states are perfectly distinguishable.● There is a single measurement where one of the
outcomes occurs with certainty on one of the states, and with probability 0 for the other
151
The gbit (one half of a PR box)
● Let us consider a system which can be prepared in 4 pure states, and mixtures of these 4 pure states.
● Its state space is the square.● Like all non-classical state spaces, there are
certain ensembles which are indistinguishable.
152
Indistinguishable ensembles
A B
CD
153
How to “read” a convex state space
A B
CD
154
How to “read” a convex state space
A B
CD
Can perfectly distinguish A,B from C,D
155
How to “read” a convex state space
A
Can we perfectly distinguish A from B?
B
156
How to “read” a convex state space
A
B
Can we perfectly distinguish A from B?
157
How to “read” a convex state space
A
B
Can we perfectly distinguish A from B?
158
How to “read” a convex state space
A
B
Can we perfectly distinguish A from B?
159
Composite systems
● As well as composing devices in sequence
● Can compose the in parallel
Preparation choice Measurement readoutMeasurement choice
160
Parallel composition
A
B
161
Operational probabilistic theories
162
Composite systems
A
B
163
Composite systems (no-signalling)
A
B
164
Composite systems (no-signalling)
B
A
165
Reduced states
A
B
166
Composite systems: state spaces
A ABB
167
Product states
A ABB
P:
168
Product states: quantum theory
169
Product states
● This map is bilinear (due to mixing)● If we assume local tomography it is just a tensor
product● But not in general
170
Reduced states
AAB
R:
171
Reduced states: quantum theory
172
Reduced states
● This map is a linear map from to● Obtained by taking the unit effect on subsytem B
and the identity on A●
173
Effect space
● We have equivalent maps on the effect space.
174
Conclusion
● Have a framework within which we can talk about operational theories: general probabilistic theories
● Highlighted the convex structure, which arises from the possibility of taking mixtures
● Mentioned the categorical structure which arises from composing devices
● Arbitrary convex sets correspond to non-classical systems
175
Redundancy of the measurement postulates of quantum theory
176
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Quantum theory
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
Alternative probabilistic structure
Outcome Probability Function
Alternative probabilistic structure
Measurement
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
Theories with modified measurements
● A measurement postulate is a set of OPFs ● for every ● Any subset which sum to 1 form a measurement● In the case of quantum theory we have
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
Probabilistic structure
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
Probabilistic structureDynamical
Structure
1) States
2) Transformations
3)
4)
5) Composition
Quantum theory
Warning!
Changing the measurement postulates will change the structure of mixed states
What are mixed states?
Mixed states
Mixed states are equivalence classes of indistinguishable ensembles.
Indistinguishability is relative to the available measurements.
Changing the measurements will change which ensembles are indistinguishable.
State spaces
Systems with alternative measurement postulates have same pure states but different mixed states.
191
Mixed states: rebit
192
Mixed states
● If we allowed different measurements, these two ensembles may be distinguishable.
193
Alternative rebit
194
Alternative rebit
195
Alternative rebit
196
Alternative rebit
197
Alternative rebit
198
Mixed states
● We can no longer represent mixed states by density matrices when we change the measurements.
● Need to re-derive the mixed states
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
200
Deriving the state space
● Find a set of fiducial measurements in
● Quantum theory
201
Deriving the state space
202
Mixed states
● The pure states of quantum theory correspond to rays in
● Every embedding of these pure states into gives a system with same pure states but different measurements
● So pure state structure of quantum theory does not fix the measurements.
203
Classifying all alternatives to the Born rule
● So can classify all alternatives to the measurements postulates
● All state spaces with same pure states but different mixed states as quantum theory
● All possible embeddings of (as a manifold) in
204
Representation theory and the Bloch sphere
205
Probabilistic representation of transformations
206
Probabilistic representation of transformations
Representation of SU(d)
207
● Alternative measurement postulate implies different mixed states
● Different representations act on these different mixed states
● Alternative measurement postulate in correspondence with representation of SU(d)
208
1) States
2) Transformations
3) Measurements
4) Probabilities
5) Composition
Theories with modified measurements
210
Composite systems
● Systems A and B and AB.●
●
● Measurement on system A and B should correspond to a joint measurement on AB.
211
Joint measurements
A
B
212
Joint measurements
AB
213
What about composite systems?
● Need a product between systems, and specifically between the OPFs:
● This product needs to obey certain constraints
● Other constraints also follow from operational considerations.
214
The quantum case:
215
Associativity
A
B
C
216
Associativity
A
B
C
217
Associativity
A
B
C
218
Associativity
● Subjective groupings of devices lead to the same operational predictions
● “pre-operational” constraint
219
What about composite systems?
220
The result
● The only OPFs which obey this associativity constraint are the quantum ones.
● Under the assumption that (as a vector space) these sets are finite dimensional.
● Proof makes use of the linear action of the unitary group on
221
The theorem
The only measurement postulate satisfying the “possibility of state estimation” has OPFs and
-product of the form
for all and where the operator satisfies and analogously for G.
222
Conclusion
● The only GPT with the same dynamical structure as quantum theory is quantum theory itself
● The measurement postulates are redundant (assuming the operational framework)
223
Other approaches
224
Other approaches
225
Conclusion
226
Perspectives on quantum theory
● Quantum logic● General probabilistic theory● Many others: Qubism, process theories...● Also many non-operational approaches
227
How to study quantum theory
● We have seen how to study GPTs and non-classical logics.
● Quantum theory is just one instance of a GPT or NCL
● What about other (non operational) approaches?● What is a general branching theory for instance?
228
Perspectives on quantum theory
● Different approaches provide different insights● In the case of studying postulates I have used two
different approaches: Quantum Logic and GPTs● We were able to compress the postulates in two
directions.● Is there a more succinct formulation of quantum
theory?
229
Perspectives on quantum theory
● There are multiple mathematical structures in quantum theory
● A lattice structure (of subspaces)● A convex structure (of state and effect spaces)● Other structures to be explored: categorical...
230
Perspectives on quantum theory
● There is a unique logic with the same measurement structure as quantum theory: quantum theory
● There is a unique GPT with the same dynamical structure as quantum theory: quantum theory
231
Perspectives on quantum theory
● There are redundancies within the postulates of quantum theory.
● But need a background framework within which we can define things and show these redundancies.